1. Field of the Invention
The present invention relates to a MOS sensor and its drive method. More specifically, the invention relates to a technique for determining the optimum storage period in a MOS area sensor.
2. Related Art
In recent years, information devices such as the personal computers have spread so widely as to enhance the desire to read various informations as electronic informations in the personal computers. Therefore, the digital still cameras have been highly noted in place of the silver salt cameras of the prior art, and the scanners have also been highly noted as the means for reading prints from paper.
In the digital still cameras, there has been used the area sensor in which the pixels of an image sensor unit are two-dimensionally arrayed. In the scanners or copying machines, there has been used the line sensor in which the pixels of the image sensor unit is linearly arrayed.
In these image reading devices, the CCD sensor is mainly used as the image sensor. In the CCD sensor, the photodiode of each pixel performs the photoelectric conversion, and its signal is read by using the CCD. In the recent years, however, the MOS sensor made of a single-crystalline silicon substrate has shown the symptom of partial spreading because a peripheral circuit can be built in, it can be made into one chip, it is suitable for the real time signal processing, and its power consumption is low, and so on. These spreading situations of the MOS sensor have been described on pp. 119 of “Nikkei Electronics”, Jul. 14, 1997 (Revived MOS Solid Image Pickup Element—Spreading Its Applications with Arms of Low Power Consumption). At the research level, however, there has also been developed the MOS sensor which is made of TFTs over a glass substrate. In these MOS sensors, the signals of individual pixels are read out by the photoelectric conversions through the photodiodes of each pixel and by the switches formed of the MOS transistors.
First of all, therefore, here will be schematically described the MOS sensor, followed by the CCD sensor.
As the pixel construction of the MOS sensor, there have been developed a variety of types. These types can be coarsely classified into two kinds: a passive sensor and an active sensor. The passive sensor is a sensor having no signal amplifying element on each pixel, whereas the active sensor is a sensor having a signal amplifying element on each pixel. The active sensor has a merit of a higher strength against noises than the passive sensor because the signal is amplified in each pixel.
FIG. 3 shows a circuit example of a pixel in the passive sensor. The pixel 305 is constructed to include a switching transistor 301 and a photodiode 304. This photodiode 304 is connected with a power reference line 306 and the source terminal of the switching transistor 301. The switching transistor 301 is connected at its gate terminal with a gate signal line 302 and at its drain terminal with a signal output line 303. The photodiode 304 performs the photoelectric conversion. Specifically, the photodiode 304 generates an electric charge according to an incident light and stores it therein. By controlling the gate signal line 302, moreover, the switching transistor 301 is turned conductive so that the electric charge of the photodiode 304 is read out via the signal output line 303.
In the passive sensor, the switching transistor 301 is turned conductive to read the charge of the photodiode 304 as the signal. Moreover, the photodiode 304 is connected with the source terminal of the switching transistor 301. When the signal is read, therefore, the photodiode 304 changes its state (e.g., the charge or the potential). In other words, the signal of the photodiode 304 cannot be read without changing its state. Therefore, it can be said that the reading by the passive sensor is a destructive one.
Also, it is ordinary that the signal of the pixel is read and outputted for each row or pixel. On the other hand, the storage period has to be equal for all pixels. It is, therefore, necessary to perform the resetting for each row or pixel. As a result, the timings for the resetting and the signal outputting are different for the individual rows (or pixels). Therefore, the storage periods are equal for all rows, but the storage timings are not. With these restrictions, however, the signals of the individual rows (or pixels) can be randomly read out on principle. Therefore, this reading method is called the “random reset/read method”.
There are a variety of types of constructions of the pixels of the active sensor. On pp. 17 (CMOS Image Sensors, Electronic Camera On a Chip) of IEDM95 or on pp. 201 (CMOS Image Sensors—Recent Advances and Device Scaling Considerations) of IEDM97, there are introduced the pixel constructions and actions of photo diode type and photo gate type. On pp. 180 (A ¼ Inch 330 k Square Pixel Progressive Scan CMOS Active Pixel Image Sensor) of ISSCC97, the pixel construction is classified from the viewpoint of a pixel selecting method. Specifically, there are described the case of using transistors and the case of using capacitors. Thus, there are various numbers of transistors which construct one pixel. In JIEC Seminar: Survey of Development of CMOS Cameras of Feb. 20, 1998, there has been widely introduced the general view of the CMOS sensors, and there has also been described the logarithmic conversion type for outputting logarithmic signals of the optical intensity by connecting the gate electrodes and the drain electrodes of resetting transistors.
The pixel construction of the active sensor, as most frequently adopted, is of the type in which one pixel 408 is composed of three N-channel transistors and one photodiode, as shown in FIG. 4. A photodiode 404 is connected at its P-channel terminal with a power reference line 412 and at its N-channel terminal with the gate terminal of an amplifying transistor 406. The drain terminal and the source terminal of the amplifying transistor 406 are connected with a power line 409 and the drain terminal of a switching transistor 401. This switching transistor 401 is connected at its gate terminal with a gate signal line 402 and at its source terminal with a signal output line 403. A resetting transistor 407 is connected at its gate terminal with a reset signal line 405. The source terminal and the drain terminal of the resetting transistor 407 are connected with the power line 409 and the gate terminal of the amplifying transistor 406.
In the case of the area sensor, one signal output line 403 is connected with not only one pixel 408 but also many pixels. However, only one biasing transistor 411 is arranged for one signal output line 403. The biasing transistor 411 is connected at its gate terminal with a bias signal line 410. The source terminal and the drain terminal of the biasing transistor 411 are connected with the signal output line 403 and a bias side power line 413.
Here will be described the basic actions of the pixel 408.
At first, the resetting transistor 407 is turned conductive. The photodiode 404 is electrically connected at its P-channel terminal with the power reference line 412 and at its N-channel terminal with the power line 409. The power reference line 412 has a potential at the reference potential of 0 V, and the power line 409 has a potential at a power potential Vdd. Therefore, a reverse bias voltage is applied to the photodiode 404. From now on, the action, in which the potential of the N-channel terminal of the photodiode 404 is charged to the potential of the power line 409, will be called the “reset”. After this, the resetting transistor 407 is turned nonconductive. Then, an electric charge is generated by the photoelectric conversion when the photodiode 404 is irradiated with light. As the time elapses, therefore, the potential, as charged to the potential of the power line 409, at the N-channel terminal of the photodiode 404 is gradually lowered due to the charge generated by the light. After a lapse of a predetermined time, moreover, the switching transistor 401 is turned conductive. Then, the signal is outputted through the amplifying transistor 406 to the signal output line 403.
When the signal is outputted, however, the potential is applied to the bias signal line 410 so that the electric current flows in the biasing transistor 411. Therefore, the amplifying transistor 406 and the biasing transistor 411 act as the so-called “source-follower circuit”.
In FIG. 4, the wiring line connecting the P-channel terminal of the photodiode 404, i.e., the power reference line 412 may be called the “photoelectric conversion element side power line”. The potential of this photoelectric conversion element side power line changes with the direction of the photodiode 404. In FIG. 4, the photoelectric conversion element side power line is connected with the P-channel terminal of the photodiode 404 and has the reference potential of 0 V. In FIG. 4, therefore, the photoelectric conversion element side power line is called the “power reference line”.
Likewise, in FIG. 4, the wiring line connecting the resetting transistor 407, i.e., the power line 409 may be called the “reset side power line”. The potential of this reset side power line changes with the direction of the photodiode 404. In FIG. 4, the reset side power line is connected through the resetting transistor 407 with the N-channel terminal of the photodiode 404 so that it has the power potential Vdd. In FIG. 4, therefore, the reset side power line is called the “power line”.
To reset the photodiode 404 is to apply the reverse bias voltage to the photodiode 404. According to the direction of the photodiode 404, therefore, there changes the magnitude relation of the potential between the photoelectric conversion element side power line and the reset side power line.
Next, FIG. 5 shows an example of the most basic source follower circuit. In FIG. 5, there is shown the case in which the N-channel transistor is used. The source follower circuit could be constructed by using the P-channel transistor. An amplification side power line 503 is fed with the power potential Vdd. A bias side power line 504 is fed with the reference potential of 0 V. An amplifying transistor 501 is connected at its drain terminal with the amplification side power line 503 and at its source terminal with the drain terminal of a biasing transistor 502. The source terminal of the biasing transistor 502 is connected with the bias side power line 504. The biasing transistor 502 is fed at its gate terminal with a bias potential Vb. Therefore, a bias current Ib flows into the biasing transistor 502. The biasing transistor 502 basically acts as a constant current source. The gate terminal of the amplifying transistor 501 acts as an input terminal 506. Therefore, the amplifying transistor 501 is fed at its gate terminal with an input potential Vin. The source terminal of the amplifying transistor 501 acts as an output terminal 507. Therefore, the source terminal of the amplifying transistor 501 takes an output potential Vout. The input/output relation of the source follower circuit at this time is Vout=Vin−Vb.
Comparing FIG. 4 and FIG. 5, the amplifying transistor 406 corresponds to the amplifying transistor 501. The biasing transistor 411 corresponds to the biasing transistor 502. It can be thought that the switching transistor 401 is omitted from FIG. 5, because the conductive state is imagined. The potential at the N-channel terminal of the photodiode 404 corresponds to the input potential Vin (i.e., the gate potential of the amplifying transistor 501 or the potential at the input terminal 506). The potential at the signal output line 403 corresponds to the output potential Vout (i.e., the source potential of the amplifying transistor 501 or the potential at the output terminal 507). The power line 409 corresponds to the amplification side power line 503.
In FIG. 4, therefore, a relation of Vout=Vpd−Vb is deduced by setting the potential at the N-channel terminal of the photodiode 404 to Vpd, by setting the potential of the bias signal line 410, i.e., the bias potential to Vb, by setting the potential of the signal output line 403 to Vout and by setting the potentials of the power reference line 412 and the biasing power line 413 to 0 V. When the potential Vpd at the N-channel terminal of the photodiode 404 changes, therefore, the potential Vout also changes so that the change in the potential Vpd can be read as the signal to read the optical intensity.
Next, the signal timing chart at the pixel 409 is illustrated in FIG. 6. At first, the resetting transistor 407 is turned conductive by controlling the reset signal line 405. Then, the potential at the N-channel terminal of the photodiode 404 is charged up to the power potential Vdd or the potential of the power line 409. In short, the pixel is reset. Then, the reset signal line 405 is controlled to turn the resetting transistor 407 nonconductive. After this, the photodiode 404 generates the electric charges according to an optical intensity if irradiated with light. Therefore, the electric charge stored by the resetting action is gradually released. In other words, the potential at the N-channel terminal of the photodiode 404 becomes lower. Where a dark light is irradiating, the discharging rate is also low so that the potential at the N-channel terminal of the photodiode 404 does not become so low. Where a bright light is irradiating, the discharging rate is so high that the potential at the N-channel terminal of the photodiode 404 drops at a high changing rate.
At an instant of time, the switching transistor 401 is turned conductive to read the potential at the N-channel terminal of the photodiode 404 as the signal. This signal is proportional to the intensity of the light. Then, similar actions are repeated by turning the resetting transistor 407 conductive again to reset the photodiode 404.
Where a very bright light is irradiating, however, the electric charge of the photodiode 404 is released so much that the potential at the N-channel terminal of the photodiode 404 becomes to a very low level. However, the potential at the N-channel terminal of the photodiode 404 does not become lower than the potential at the P-channel terminal of the photodiode 404, i.e., the potential of the power reference line 412. When an intense light irradiates, therefore, the potential at the N-channel terminal of the photodiode 404 becomes low. When this potential becomes lower and lower down to the potential of the power reference line 412, it does not change any more. This situation is called the “saturation”. In this saturation, the potential at the N-channel terminal of the photodiode 404 does not change so that the correct signal, i.e., the signal according to the optical intensity cannot be outputted. Within the normal action range, therefore, it is necessary to prevent the photodiode 404 from being saturated.
Here, the period from the time when the pixel is reset to the time when the signal is outputted is called the “storage period”. In short, the storage period is the time period, for which the signal is being stored by irradiating the light receiving unit of an image sensor with the light, and is called the “storage term” or the “exposure period”. For the storage period, the photodiode 404 is storing the electric charge generated by the light. For the different storage periods, therefore, the totals of the electric charges, as generated with the light, are different even for an identical light intensity, so that the signal values become different. For example, a bright light causes a saturation for a short storage period. Even a dark light will cause a saturation if the storage period is long. In short, the signal is determined by the product of the optical intensity and the storage period.
FIG. 6 illustrates the case of one pixel. Here will be described the case in which the pixels are arrayed in a matrix shape. In this case, the signals of the pixels are read out as outputs for each row. On the other hand, the storage period has to be equal for all pixels. Therefore, the resetting has to be done for each row. As a result, the timing for the resetting and the timing for outputting the signals are different for the individual rows. Therefore, the storage period is equal for the pixels of all rows, but the storage times are different. With these restrictions, however, the signals of the individual rows can be read out at random. Therefore, the reading method of this case is the random reset/read method.
In the pixel 408, on the other hand, the switching transistor 401 is turned conductive to read the signal from the photodiode 404. However, the N-channel terminal of the photodiode 404 is connected with the gate terminal of the amplifying transistor 406. Even if the signal is read, therefore, no change occurs in the state (e.g., the charge or potential) of the photodiode 404. In other words, the signal of the photodiode 404 can be read many times without changing the state of the photodiode 404. Therefore, it can be said that the reading in the active sensor is a non-destructive reading.
Here will be described the transistors in the pixel 408. On the polarity, the transistors are frequently of the N-channel type. It is rare, but the resetting transistor is of the P-channel type (as should be referred to FIG. 11 of pp. 9 of JIEC Seminar: Survey of Development of CMOS Cameras, Feb. 20, 1998). On the other hand, the amplifying transistor and the selecting transistor, which are both N-channel types, are frequently arranged, as shown in FIG. 4, by connecting the power line 409 and the amplifying transistor 406, by connecting the amplifying transistor 406 and the switching transistor 401, and by connecting the switching transistor 401 and the signal output line 403. The arrangement is rarely effected by using both the transistors of the N-channel type to connect the power line 409 and the switching transistor 401, the switching transistor 401 and the amplifying transistor 406, and the amplifying transistor 406 and the signal output line 403 (on pp. 180 of ISSCC97: A ¼ Inch 330 k Square Pixel Progressive Scan CMOS Active Pixel Image Sensor).
The MOS sensor has been described hereinbefore. Here will be described CCD sensor.
At first, the CCD sensor can be coarsely classified into two in accordance with the signal transfer method in the CCD. One is the frame transfer type CCD, and the other is the interline transfer type CCD. Basically, both the CCDs act in the bucket relay method. Specifically, a signal is transferred to an adjoining pixel, and another signal is received from another adjoining pixel. These actions are repeated. These actions are made for all pixels and are repeated to transfer the whole signals. Only a signal of a certain pixel cannot be read by itself but is transferred to an adjoining pixel at all times.
The frame transfer CCD shares the light receiving unit and the signal transferring unit. In the actions, the light receiving unit performs the photoelectric conversion to store the electric charge and transfers the signal to an adjoining light receiving unit (i.e., transfers a signal to an adjoining pixel and receives another signal from another adjoining pixel). In this method, an electric charge is mixed, if generated by a new light while the signal is being transferred, into the signal being transferred. This makes it necessary to shut the light while the signal being transferred. Therefore, a mechanical shutter is used to shut the light.
In the interline transfer CCD, the transfer CCD is arranged separately from the light receiving unit. The action is to transfer the signals, which are stored in the light receiving unit, all at once to the transfer CCD. After this, the transfer CCD transfers the signals (i.e., transfers a signal to an adjoining pixel and receives another signal from another adjoining pixel). The interline transfer CCD is shut from the light, and the light receiving unit and the CCD unit are separated. Even if the light is produced in the light receiving unit while the signal is being transferred, therefore, it is not mixed into the signal being transferred. Therefore, no problem arises even if the light is irradiating while the signal is being transferred.
The timing of the resetting of the CCD sensor is different from that of the case of the MOS sensor. In the case of the MOS sensor, the signals are read out one by one or row by row from the pixels so that the resetting is done one by one or row by row. In the CCD sensor, on the other hand, the reading is started all at once from all pixels. In order to equalize the storage periods, therefore, it is necessary to reset the pixels all at once. As a result, all the pixels are reset simultaneously and the singles are outputted from all the pixels simultaneously. Therefore, the storage periods are equal for all pixels, and the storage times are identical. Thus in the CCD sensor, the pixels are reset all at once and are read all at once. Therefore, this reading method is called the batch reset/read method.
In the CCD sensor, on the other hand, the electric charge, as stored by the light receiving unit, is transferred. After the signal has been once read, the state (i.e., the charge or potential) of the light receiving unit changes. In other words, the signal of the light receiving unit cannot be read without changing the state of the light receiving unit. Therefore, it can be said that the reading in the CCD sensor is a destructive reading.
On pp. 47 (CCD for Digital Camera Steered from Exclusive Devotion to Pixel Number to Improvement in Sensitivity) of Nikkei Electronics (No. 732), Dec. 14, 1998, and on pp. 159 (Appearance of CCD Directed to Personal Computer Camera) of Nikkei Electronics (No. 634), Apr. 24, 1995, there is introduced the method of the CCD sensor. On pp. 261 of Nikkei Electronics (Solid Image Pickup Element Cameras Having Been Reported These 18 Years) of Nikkei Electronics, Sep. 14, 1992, there have been introduced differences between the MOS sensor and the CCD sensor.
Here will be described the sensor unit for the photoelectric conversion. The CCD sensor and the MOS sensor have no special difference in the sensor unit. Usually, a PN type photodiode is used to convert a light into electricity. The sensor unit is further exemplified by a PIN type diode, an avalanche diode, an npn buried type diode or a Schottky diode. Another sensor unit may be a photoconductor for an X-ray or a sensor for an infrared ray. This sensor unit is described in “Fundamentals of Solid Image Pickup Element—Device of Electronic Eyes” of Nippon Rikoh Shuppankai written by Takao Ando and Hirohito Komofuchi.
Here will be described appliances suited for the sensor. This sensor is used not only in the ordinary digital still camera or scanner but also in the X-ray camera. This camera may use a photoconductor for converting the X-ray directly into electric signals or may read the light which has been converted from the X-ray by a fluorescent material or a scintillator. On pp. 203 of Euro Display 99 (X-ray Detectors based on Amorphous Silicon Active Matrix), there has been described the case in which the X-ray is converted into the light by the scintillator and the light is read out. On pp. 21 of IEDM 98 (Amorphous Silicon TFT X-ray Image Sensors), it has been reported that the light is read by means of amorphous silicon. On pp. 45 of AM-LCD99 (Real-Time Imaging Flat Panel X-ray Detector), it has been reported that the light is read by means of the photoconductor.
Next, it will be considered what range the optical intensity of an image falls under when the object is to be read by using the CCD or MOS image sensor.
In the first case of the digital still camera, the optical intensity of the object ranges from the black state to such a bright state as experienced by observing the sun directly. Thus, the optical intensity of the object can take a range from 0 to infinity. Therefore, the image sensor to be employed here is required to have a wide dynamic range for the incident light. As a matter of fact, however, the image sensor has a limited dynamic range so that the imaging has to be done for a standard object illuminance. If the object illuminance is improper, a flash is frequently used. Alternatively, the shutter is used to adjust the exposure time. The shutter of the digital still camera has two kinds: a mechanical shutter and an electronic shutter. The mechanical shutter shuts the optical slit mechanically as in the case of the silver salt camera. The electronic shutter changes the storage period by adjusting the drive signal of the image sensor.
On the other hand, the scanner is mostly prepared with a dedicated light source. Even if the object had a reflectivity of 100%, therefore, the range of the intensity of the light to enter the image sensor is known in advance. In other words, a more intense light than that of the dedicated light source will not enter. Thus, the storage period may be so set that the output signal may be saturated where the reflectivity is the highest (as usually experienced by the white paper).
The potential at the N-channel terminal of the photodiode 404 hardly changes, when the optical intensity is low, but highly changes when the intensity is high. Where the optical intensity is extremely high, however, the potential at the N-channel terminal of the photodiode 404 drops as low as the potential at the P-channel terminal of the photodiode 404 and may be saturated. The potential at the N-channel terminal of the photodiode 404 does not change any more, if saturated, so that the image cannot be correctly read. Even in the high optical intensity, therefore, the storage period has to be so shortened for adjustment as to prevent the saturation.
If the storage period is excessively shortened, however, it may end although the potential at the N-channel terminal of the photodiode 404 changes a little. In this case, the signal amplitude is reduced to degrade the image quality.
It is, therefore, desired to prevent the saturation and to enlarge the signal amplitude even where the optical intensity is high. This desire can be satisfied if the storage period is so adjusted that the signal to be outputted may take a value just before the saturation. It is found from the description thus far made that the optimization of the storage period is important.
It is, therefore, assumed that the image is taken for examining the optimum storage period. This image pickup will be called the “trial imaging”.
In the case of the CCD sensor, the reading method is the batch reset/read method, by which the signals are transferred as if they were relayed in buckets. This method makes it impossible to reset or to read out the signals for every pixels. In other words, it is impossible to use the random reset/read method. Where the signals of one frame are read, therefore, the storage periods of all pixels are equalized. In one image pickup, the storage periods cannot be changed for every pixels. Because of the destructive reading, on the other hand, the image pickup has to be done over again after one trial.
It is assumed in the situations described above that the trial imaging is done by setting the storage period to a value and by reading the signals of one frame. In this case, however, the signals may have been saturated already. Then, the trial imaging has to be redone. Moreover, it is necessary to set the tentative storage period again. In this case, however, it is unknown what value the tentative storage period has to be set to. What is known is that the tentative storage period may be shorter than the storage period at the time of the first trial imaging. If the storage period is shorter, the signal amplitude may be so small that it cannot be correctly read out. If the storage period is still longer, the trial imaging has to be done once more.
Here will be illustrated an example of the case described above. FIG. 7 illustrates a change in the potential at the N-channel terminal of the photodiode after reset. Since the period just before the saturation is the optimum storage period, it is found from FIG. 7 that the storage period is optimum at the period 10 till the reading.
In order to find out the optimum period, a first trial imaging is done. Since the optimum storage period is absolutely unknown at this time, it is assumed that the trial imaging is done for a storage period of 20. At this time, however, it is found from FIG. 7 that the saturation has been already completed. Therefore, a second trial imaging is done for a storage period of 15. However, the saturation has been still completed. Therefore, a third trial imaging is done over again for a storage period of 8. Then, the saturation is not completed so that the signal value according to the optical intensity is outputted. The storage period is analogized from the signal value at this time.
In this example, the trial imaging is done three times. The time period necessary for the trial imaging is as long as 20+15+8=43, as shown in FIG. 8.
Thus, for the CCD sensor, it is seriously difficult to find out the optimum storage period.
Thus, there has been described the storage period for preventing the signals from being saturated. Here will be described the difficulty for the signals to be precisely read where the object to be imaged has close graduations (or brightnesses).
FIG. 9 shows a signal timing chart at the pixel 409. At first, the resetting transistor 407 is turned conductive by controlling the reset signal line 405. Then, the potential at the N-channel terminal of the photodiode 404 is charged to the power potential Vdd or the potential of the power line 409 so that the pixel is reset. Then, the resetting transistor 407 is turned nonconductive by controlling the reset signal line 405. After this, the potential at the N-channel terminal of the photodiode 404 becomes lower if the photodiode 404 is irradiated with a light. Where the photodiode 404 is irradiated with a dark light, the amount of the discharge is so little that the potential at the N-channel terminal of the photodiode 404 does not become so low. Where photodiode 404 is irradiated with the bright light, the amount of the discharge is so much that the potential at the N-channel terminal of the photodiode 404 becomes extremely low.
At a point of time, moreover, the potential at the N-channel terminal of the photodiode 404 is read out as the signal by turning the switching transistor 401 conductive.
Here, it is assumed that the photodiode 404 is irradiated with lights of similar intensities. The potentials at the N-channel terminal of the photodiode 404 in this case are so close, as illustrated in FIG. 9. Therefore, it is made difficult by the influences of noises or dispersions to detect the potential difference.